Canada’s AI Strategy Leaves Copyright Issues Unaddressed

Canada’s National Artificial Intelligence (AI) Strategy, known as ‘AI for All’, outlines how Canada will drive AI adoption, and build trust and infrastructure, with a clear policy objective of fostering a sovereign and globally competitive AI ecosystem. The Strategy is generally silent on whether any changes or updates to IP law are contemplated. 

Central to the strategy is an emphasis on Canadian control over AI systems. However, from an intellectual property (IP) standpoint, it remains unclear to what extent AI-generated outputs are protectable, and whether and in what circumstances training AI using copyright protected works may violate copyright.   

These concerns extend the issues we raised in our recent article Canada’s AI Strategy Starts but Falls Short on Ensuring Patent Protection for AI Inventions, which highlights gaps in the Strategy from a patent perspective.  

In this article, we will outline how the Strategy similarly leaves lingering questions from a copyright perspective. Although Canadian government officials have indicated that copyright issues remain under active consideration and that further policy development is expected, the absence of concrete direction leaves a number of important issues unresolved and reinforces the current state of legal uncertainty. 

Key Copyright Issues 

Generally, the protectability of AI-generated outputs will remain a live issue in Canada. Where AI is prompted to generate text, music, images, or code, it remains unclear whether and in what circumstances those outputs may attract copyright protection under Canadian law. Questions of authorship and ownership in this context are unresolved and will become increasingly important as AI adoption scales, particularly given the stated goal of reaching 60% business adoption within the next eight years. 

The arguably existential question of whether, and in what circumstances, text and data mining (TDM) infringes copyright also remains outstanding. Generative AI systems rely on large-scale ingestion of data, including copyrighted works, to train AI systems, and this has already led to litigation in Canada involving allegations of copyright infringement.  In addition to future training activities, there is ambiguity regarding whether any safe harbour may emerge for AI systems trained outside of Canada but deployed domestically without further training. Liability for infringement arising from AI-generated outputs also remains unsettled, including how responsibility may be allocated among users, developers, and deployers. 

Further, the approach to transparency remains unclear. Public consultation feedback to date has underscored the importance of transparency for copyright holders, particularly in enabling them to understand how their works are used to train AI, and whether something was generated by AI. A government-initiated public consultation on transparency may follow; however, it remains uncertain whether copyright-specific considerations will be addressed.  

The government has undertaken multiple public consultations on AI. However, with copyright largely absent from the current AI Strategy, nagging copyright issues relating to copyright protection for AI-generated outputs, and whether TDM and training practices may violate copyright in works used in training, remain without clear policy direction. For now, these issues appear likely to be addressed incrementally through litigation, absent legislative reform. 

Takeaway 

In the continued absence of guidance from the Canadian government on copyright, businesses will need to manage risk in real time. 

 If you would like to discuss strategies for managing copyright risk in the development or deployment of AI systems, please contact a member of our Marks & Clerk Copyright team. 

This article is intended to provide general information only and should not be relied on as legal advice.